#!/usr/bin/env python3
"""Regenerate _to-read.md: dangling paper links across the vault.

A slug like `benjamin2019-fdr` referenced by notes but having no
papers/benjamin2019-fdr.md is a paper the discourse points to that we haven't
analyzed — the to-read list. Reference count is the in-vault influence proxy.

The section below the MANUAL marker is preserved verbatim across runs — put
titles, priorities, and notes there.

    python3 _scripts/make_toread.py
"""
import glob
import os
import re
from collections import defaultdict

VAULT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
SCAN_DIRS = ["papers", "influences", "ideas", "concepts", "syntheses", "methods"]
MARKER = "<!-- MANUAL — everything below survives regeneration -->"


def main() -> None:
    existing = set()
    for d in SCAN_DIRS + ["_templates", "authors"]:
        existing |= {os.path.splitext(os.path.basename(p))[0]
                     for p in glob.glob(os.path.join(VAULT, d, "*.md"))}
    existing |= {os.path.splitext(f)[0] for f in os.listdir(VAULT) if f.endswith(".md")}

    refs = defaultdict(set)
    idea_refs = defaultdict(int)
    for d in SCAN_DIRS:
        for path in glob.glob(os.path.join(VAULT, d, "*.md")):
            stem = os.path.splitext(os.path.basename(path))[0]
            for line in open(path, encoding="utf-8"):
                if line.startswith("authors:") or line.strip().startswith("uses::"):
                    continue  # author and trick nodes dangle by design
                clean = re.sub(r"`[^`]*`", "", line)  # code spans aren't links
                for t in re.findall(r"\[\[(.+?)\]\]", clean):
                    t = t.split("|")[0].split("#")[0].strip()
                    if t and t not in existing and re.match(r"^[a-z][a-z-]*\d{4}-", t):
                        refs[t].add(stem)
                        if d == "ideas":
                            idea_refs[t] += 0  # counted below via source sets
    # idea-refs = how many of a slug's referencing notes are idea notes
    idea_stems = {os.path.splitext(os.path.basename(p))[0]
                  for p in glob.glob(os.path.join(VAULT, "ideas", "*.md"))}
    for t in refs:
        idea_refs[t] = sum(1 for s in refs[t] if s in idea_stems)

    # External influences: registered out-of-scope slugs leave the to-read
    # queue and get their own ranked table inside _influences.md.
    inf_path = os.path.join(VAULT, "_influences.md")
    external = set()
    if os.path.exists(inf_path):
        inf_text = open(inf_path, encoding="utf-8").read()
        register = inf_text.split("## Register", 1)[1] if "## Register" in inf_text else ""
        # In-scope rulings are not external — but "Wave-N additions" blocks
        # appended after that heading are register entries; keep those lines.
        if "## Adjudicated IN scope" in register:
            register, adjudicated = register.split("## Adjudicated IN scope", 1)
            in_wave = False
            for line in adjudicated.splitlines():
                if re.match(r"^Wave-\d+ additions", line):
                    in_wave = True
                elif line.startswith("## "):
                    in_wave = False
                if in_wave:
                    register += line + "\n"
        external = {t.split("|")[0].split("#")[0].strip()
                    for t in re.findall(r"\[\[(.+?)\]\]", register)
                    if re.match(r"^[a-z][a-z-]*\d{4}-", t.split("|")[0].strip())}
        ext_refs = {s: refs[s] for s in external if s in refs}
        gen = ["", "| External root | Refs | Imported by |", "|---|---|---|"]
        for slug in sorted(ext_refs, key=lambda s: (-len(ext_refs[s]), s)):
            srcs = ", ".join(f"[[{s}]]" for s in sorted(ext_refs[slug]))
            gen.append(f"| [[{slug}]] | {len(ext_refs[slug])} | {srcs} |")
        gen.append("")
        start, end = "<!-- GENERATED:START (by make_toread.py — ref counts + importing notes) -->", "<!-- GENERATED:END -->"
        if start in inf_text and end in inf_text:
            pre = inf_text.split(start, 1)[0]
            post = inf_text.split(end, 1)[1]
            open(inf_path, "w", encoding="utf-8").write(pre + start + "\n" + "\n".join(gen) + end + post)
            print(f"wrote {len(ext_refs)} external influences to _influences.md")
        refs = {k: v for k, v in refs.items() if k not in external}

    out = os.path.join(VAULT, "_to-read.md")
    manual = ""
    if os.path.exists(out):
        prev = open(out, encoding="utf-8").read()
        if MARKER in prev:
            manual = prev.split(MARKER, 1)[1]

    lines = [
        "---",
        "tags: [meta]",
        "---",
        "",
        "# To-read list",
        "",
        "Papers the vault's discourse points to but that have no note yet —",
        "generated by `_scripts/make_toread.py` from dangling links; reference",
        "count is the in-vault influence proxy. Feed a slug to `/analyze-paper`",
        "and it disappears from this list on the next regeneration.",
        "",
        "| Wanted paper | Refs | Idea-refs | Referenced from |",
        "|---|---|---|---|",
    ]
    for slug in sorted(refs, key=lambda s: (-len(refs[s]), -idea_refs[s], s)):
        srcs = ", ".join(f"[[{s}]]" for s in sorted(refs[slug]))
        lines.append(f"| [[{slug}]] | {len(refs[slug])} | {idea_refs[slug]} | {srcs} |")
    lines += ["", MARKER]
    body = "\n".join(lines) + (manual if manual else "\n\n## Notes / titles / priorities\n\n(add freely — preserved across regenerations)\n")
    open(out, "w", encoding="utf-8").write(body)
    print(f"wrote {len(refs)} wanted papers to _to-read.md")


if __name__ == "__main__":
    main()